Responsiveness of the PedsQL to Pain-Related Changes in Health-Related Quality of Life in Pediatric Sickle Cell Disease

Abstract

Objective To determine if caregiver report of the Pediatric Quality of Life Inventory (PedsQL) is responsive to changes in health-related quality of life (HRQL) associated with pain episodes in pediatric sickle cell disease (SCD). Methods 81 caregivers of children ages 2–19 years with SCD completed the PedsQL as part of routine psychosocial screenings at 2 time points, ranging from 6 to 18 months apart. Frequency of SCD-related pain episodes between time points was assessed using medical chart review. Results The frequency of pain episodes between time points was a significant predictor of decreases in physical, psychosocial, and total HRQL, even after controlling for time interval, demographic, and medical variables. Conclusions The caregiver report of the PedsQL appears to be a useful tool for capturing changes in HRQL over time associated with pain episodes in SCD.

Introduction

A primary purpose of studying health-related quality of life (HRQL) is to identify aspects of a patient’s quality of life that are affected by a health condition and that may be amenable to change by intervention. To achieve this aim, measures of HRQL should accurately capture the changes in HRQL that occur as the result of a disease process or specific medical complication (Pal, 1996). In HRQL research, this concept is referred to as the responsiveness of a measure (Terwee, Dekker, Wiersinga, Prummel, & Bossuyt, 2003). In pediatric sickle cell disease (SCD), vaso-occlusive pain is considered a hallmark feature of the disease and the most frequent source of morbidity in patients (Platt et al., 1991). In childhood, recurrent pain is marked by both physical and psychosocial difficulties, including reduced opportunities for physical recreation, fewer social opportunities, frequent school absences, and increased reports of depression and anxiety (Anie, 2005; Edwards et al., 2005; Fuggle, Shand, Gill, & Davies, 1996).

Despite links between pain and difficulties in these specific areas of child adjustment, measures designed to assess HRQL constructs have been inconsistent in capturing reductions in physical and psychosocial well-being that would be anticipated in children with recurrent pain from SCD. Among cross-sectional studies that have examined pain and HRQL in pediatric SCD, two have found that pain is related to lower physical and psychosocial HRQL (Dampier et al., 2010; McClellan, Schatz, Sanchez, & Roberts, 2008). The remaining studies have found that pain is only related to physical HRQL or that pain has no relationship to HRQL (Palermo, Riley, & Mitchell, 2008; Palermo, Schwartz, Drotar, & McGowan, 2002; Panepinto, O'Mahar, DeBaun, Loberiza, & Scott, 2005).

These inconsistent findings may be partially explained by variations in methods used to measure pain. For example, several studies have used health care contacts as a proxy for pain; however, some have used only one form of health care contact (e.g., hospitalizations for pain) or have combined pain with other medical complications (e.g., pain and acute chest syndrome) to predict HRQL (Palermo et al., 2002; Panepinto et al., 2005; Panepinto, Pajewski, Foerster, Sabnis, & Hoffmann, 2009). These methods may underestimate the impact of pain or may obscure the relationship between pain and HRQL amidst other complications that are measured. It is also possible that HRQL measures differ in their responsiveness to pain-related effects on HRQL.

The study of HRQL is also complicated by the large number of variables that may be linked to child well-being, ranging from individual child characteristics to neighborhood/community-level influences. For example, the following variables have been associated with HRQL in pediatric SCD: child age, gender, disease severity, and history of medical comorbidities; caregiver education, income, and locus of control; and distressed neighborhood (Barakat, Lutz, Smith-Whitley, & Ohene-Frempong, 2005; Palermo et al., 2002, 2008; Panepinto et al., 2005, 2009). The inclusion of these variables in models of HRQL may make it difficult to isolate the effects of pain due to shared variability between pain and related constructs (e.g., medical comorbidities) and pain may not necessarily be the focus of HRQL studies.

In addition to inconsistencies found in cross-sectional studies of HRQL, there has been limited research on whether HRQL measures are responsive to changes in pain over time in SCD. This information is particularly important for guiding the use of HRQL measures to monitor patient progress and response to treatment for pain over time (Dampier et al., 2010; Terwee et al., 2003). Only one study has examined HRQL in pediatric SCD in a prospective manner to examine pain-related changes. Using the Acute Version of the PedsQL, Brandow, Brousseau, Pajewski, and Panepinto (2009) studied short-term changes in caregiver- and child-reported HRQL in 57 children admitted into the emergency room for pain. This research group found immediate declines in both physical and psychosocial HRQL, with ratings improving to the levels of SCD controls 1 week after the visit. Although this study suggests that the PedsQL is responsive to immediate and short-term pain-related changes in HRQL, additional research is needed to understand the responsiveness of measures over longer time intervals. In particular, using periods of time that are characteristic of other situations in which HRQL might be assessed (e.g., during routine health maintenance visits, pre and postintervention) would be particularly useful.

The aim of the present study was to examine the responsiveness of caregiver reports of the PedsQL to pain-related changes in physical, psychosocial, and overall HRQL. Strengths of this study included the use of multiple forms of health care contact to assess pain and the use of a prospective approach to examine change over time. Based on the only previous study of change over time by Brandow et al. (2009), we hypothesized that children who experienced pain during the intermediary of two time points would experience reduced physical, psychosocial, and overall HRQL over time based on caregiver report.

Methods

Participants

This study used previously collected data from a routine monitoring program for psychosocial well-being in SCD. Medical chart reviews were conducted on 94 children with SCD between the ages of 2 and 19 years who participated in psychosocial screenings between September of 2004 and January of 2008 at a hospital-based Pediatric Hematology/Oncology outpatient clinic in Columbia, South Carolina. Participation in routine psychosocial screenings as part of health maintenance visits at the clinic was strongly encouraged by the attending hematologist and was considered part of the child’s care at the clinic. Families were approached for an initial screening during routine visits. They were approached a second time at their next scheduled routine visit. No families refused to participate in the screenings; however, some families could not participate due to time constraints and were approached at a subsequent visit. Information from children and caregivers who completed the screenings at Time 1 was used in a previous publication that assessed the psychometric properties of the PedsQL in pediatric SCD; including internal consistency reliability, construct validity, criterion validity, and comparisons of child and caregiver ratings (McClellan et al., 2008).

Table I provides descriptive information on the caregivers who participated in the screenings at each visit and their children with SCD. All families identified as African-American. Although both caregivers and children reported on HRQL during the screenings, this study focuses on caregiver reports due to higher rates of completion of HRQL forms across time points. Of the total 94 families, 90 caregivers had ratings at Time 1. Eighty-one caregivers had ratings at both time points and their data was included in the present study. Nine caregivers were lost to follow-up at Time 2 with no documented reason. At both time points, some caregivers were unavailable to complete HRQL forms due to 18- to 19-year-old children completing health visits independently. Four caregivers were unavailable at Time 1 and three were unavailable at Time 2. One caregiver had Time 2, but not Time 1 HRQL ratings.

Table I.

Descriptive Information for Caregivers and Their Children With Sickle Cell Disease

Information type	Time 1	Time 2
Time between ratings (M, SD)	–	12.36 (3.24)
Total families at screenings (n)	94	85
Caregivers with HRQL ratings (n)	90	81
Caregiver type (n)
Mother	79	71
Father	7	7
Guardian	4	3
Child age at Time 1 (M, SD)	11.04 (4.34)	11.09 (4.40)
Child gender
Male/Female	44/46	39/42
Sickle cell subtype
HbSS	64	56
HbSC	13	12
HbSβ+	6	6
HbSβ0	7	7
Hematocrit at time 1 (M, SD)	28.16 (5.17)	28.19 (5.30)
Insurance status
Medicaid only	46	40
Medicaid and private insurance	16	16
Private insurance only	28	25
History of neurobehavioral complications (n)
Present/Absent	49/41	46/35
History of medical complications (n)
Present/Absent	61/29	53/28
Pain episodes between time points (M, SD)	–	1.59 (3.36)

Note. Medical information is for children with sickle cell disease whose HRQL was reported by caregivers. Neurobehavioral complications included: history of stroke, silent cerebral infarcts, moya moya disease, meningitis, seizure disorder, or formally diagnosed disorders in learning, language, or development. Medical complications included: history of acute chest syndrome, splenic sequestration, sepsis, aplastic crisis, priaprism, pneumonia, tonsillectomy/ adenoidectomy, and cholecystectomy. Pain episodes included documented hospitalizations, emergency room visits, and contacts or visits to the clinic for pain.

During the time period of the study, an estimated 115 families with children of ages 2–19 years accessed services at the clinic. Thus, our Time 1 sample of caregivers represented approximately 78% of the clinic population and our final sample of caregivers with both Time 1 and Time 2 ratings represented 70% of the clinic population. The psychologist attempted to make contact with as many families as possible, and we suspect that many of the families who did not participate or who were lost to follow-up were unable to attend routine visits with the hematologist or were completing health maintenance visits elsewhere.

Measures

PedsQL

Caregivers reported on HRQL using the Pediatric Quality of Life Inventory (PedsQL), Fourth Edition at both time points (Varni, Seid, & Kurtin, 2001). The caregiver proxy form for HRQL was designed for children of ages 2–18 years. In the present study, two participants were 18 years old at Time 1 and turned 19 years prior to Time 2. The inclusion of these participants did not alter the results of the study and they were retained in the statistical analysis. The PedsQL includes 23 items that assess the child’s functioning in four domains: Physical (eight items), Emotional (five items), Social (five items), and School (five items). Caregivers indicate how much of a problem the child has had over the previous month with specific aspects of functioning on a 0- to 4-point scale (0 = never a problem; 4 = almost always a problem). The items are linear transformed to a 0–100 scale with higher scores indicating higher HRQL. Domain scores are derived by adding up the scores on the items within the domain and dividing by the number of items answered. The Emotional, Social, and School domains are averaged to obtain a Psychosocial domain score. The Total Score (a summary of all domains) is derived by summing all item scores and dividing the score by the number of items answered. If less than 50% of the items for any domain are missing, the scale score is not calculated; otherwise, domains with missing items are prorated.

The PedsQL has been used previously with caregivers of children with SCD. As reported by Panepinto, Pajewski, Foerster, and Hoffman (2008), internal consistency for caregivers is adequate for the Physical (α = .91), Psychosocial (α = .89), and Total Score HRQL (α = .93). A previous publication based on the present sample at Time 1 also suggested adequate internal consistency for Physical (α = .82) and Total Score HRQL (α = .89) (McClellan et al., 2008). The caregiver report of the PedsQL has demonstrated criterion validity in that it is able to differentiate between specific subgroups of children with SCD (e.g., children with and without pain or neurobehavioral complications), children with mild versus severe SCD, and children with and without SCD (Panepinto et al., 2008). Convergence between caregiver and child ratings on the PedsQL has ranged from poor to very good agreement. Children under the age of 13 years may exhibit lower agreement (r = −.02 for Physical HRQL and r = .38 for Psychosocial HRQL) with caregivers versus children over the age of 13 (r = .43 for Physical HRQL and r = .51 for Psychosocial HRQL) (McClellan et al., 2008).

Medical Chart Review

Patient demographics and history of disease complications were obtained through medical chart review of paper and electronic records using a structured medical chart checklist. As reported in a previous publication by the authors, κ-coefficients ranged from .64 to .85, representing “good” to “very good” inter-reliability estimates (Fleiss, 1981; McClellan et al., 2008).

Demographic information included caregiver relationship to the child and child gender, ethnicity, age, and insurance status (Medicaid only, combined Medicaid and private insurance, or private insurance only). Insurance status was used as a proxy for income, with Medicaid only representing lower income, combined insurance representing middle income, and private insurance only representing higher income. Disease-specific information included child history of neurobehavioral complications (history of stroke, silent cerebral infarcts, moya moya disease, meningitis, seizure disorder, or formally diagnosed disorders in learning, language, or development), and history of SCD-related medical complications (history of acute chest syndrome, splenic sequestration, sepsis, aplastic crisis, priaprism, pneumonia, tonsillectomy/adenoidectomy, and cholecystectomy). We also collected information on hematocrit, which is a biomarker of disease severity (Platt et al., 1991).

History of health care contacts for pain between Time 1 and Time 2 (number of hospitalizations, emergency room visits, and contacts or visits to the clinic for pain) was used to determine pain episode frequency. It is standard practice for the hematologist to ask families about pain episodes, including those treated at outside facilities, at every routine visit and for medical personnel to document when the presenting concern of patients is pain from SCD or another problem (e.g., acute chest syndrome). Although we do not have specific rates of utilization for pain versus other complications at our clinic, we suspect that pain is the primary reason for health care utilization based on previous studies of patients with SCD. Healthcare utilization when pain is a symptom has been used a proxy for pain frequency in several previous studies assessing pain episode frequency in SCD (Ballas et al., 2009; Platt et al., 1991).

Procedures

Institutional Review Board approval was obtained concurrently from Palmetto Richland Hospital and the University of South Carolina prior to conducting the medical chart review. A passive consent process was employed to allow families the opportunity to opt out of the study. Families were mailed a letter describing the study and were asked to return the letter if they preferred not to have their child’s medical record reviewed. No letters were returned. Medical chart reviews and screenings were conducted by the primary investigators of this study (C.B.M. and J.S.), both psychologists who were familiar with the clinic structure. Electronic and paper medical records were obtained for families who had participated in the screenings from September of 2004 to January of 2008. Both types of chart records were reviewed to ensure that documents from outside treating facilities were included in the review. Records were de-identified at the time of the chart review to ensure confidentiality.

Psychosocial screenings were conducted on a regular basis at the clinic as part of routine hematological care, with most children receiving screenings annually. After the child’s physical examination, families were accompanied by the hematologist to the psychologists’ office located in the clinic to complete the screening. The psychologist explained that the purpose of the screening was to check on all aspects of a child’s well-being and to introduce the psychosocial services available at the clinic. Caregivers were initially queried as to current concerns in various areas of child development (e.g., routines, behavior, family life) using a semi-structured interview. Next, the PedsQL was introduced as a tool to provide a better understanding of the child’s functioning in their daily life. The psychologist reviewed the instructions of the PedsQL and demonstrated how to complete a few sample items to ensure comprehension.

Information from the semi-structured interviews and HRQL forms were used to provide referrals to psychosocial services available at the clinic. Referrals were made to families if the caregiver indicated concerns in a particular area or if the psychologist identified areas of potential intervention. For the final sample of 81 families, at Time 1, the following referrals were made: neuropsychological or psycho-educational evaluation (n = 37), counseling (n = 17), school advocacy (n = 23), pain management (n = 2), developmental screening (n = 2), pill swallowing training (n = 1), and social support (n = 1).

Families were approached again at their next routine visit (ranging from 6 to 12 months later, depending on the severity of the child’s illness and disease subtype). Families who could not complete the screening at this visit were approached again at the child’s next appointment. The time between screenings ranged from six to eighteen months (M = 12.36, SD = 3.24). Screenings at the second time point were similar to the initial appointment and involved a semi-structured interview and report of HRQL using the PedsQL.

Data Analysis

Three multiple hierarchical regressions were used to examine changes in physical, psychosocial, and overall HRQL that could be predicted by pain episode frequency (i.e., number of health care contacts for pain) between time points. Bivariate correlations between the change in HRQL (using standardized residual scores) and potential confounding variables, including child age (continuous), gender (categorical), insurance status (categorical), history of neurobehavioral comorbidities (categorical), and history of medical complications (categorical) were conducted to determine their inclusion in the model. Variables were selected based on previous research demonstrating relationships between these variables or similar constructs and HRQL in pediatric SCD (Palermo et al., 2002; Panepinto et al., 2005, 2009).

Physical, Psychosocial, and Total HRQL ratings of the PedsQL at Time 2 were the dependent variables for the regressions. In all three analyses, the first step contained Time 1 HRQL ratings (Physical, Psychosocial, or Total). Step two contained the time interval (continuous) between screenings to control for variation in time between ratings. Step three contained confounding demographic variables, such as age and gender. Step four contained confounding medical variables, such as history of neurobehavioral and medical complications. The final step contained the number of health care contacts for pain (i.e., pain episode frequency; continuous) between the two time points. A bonferonni correction was used to correct for multiple comparisons. An α-level of .05 was divided by three comparisons for Physical, Psychosocial, and Total HRQL resulting in an adjusted α of p < .016.

We were unable to generate an accurate expected effect size to calculate power and sample size estimates. Previous studies have largely examined pain in a cross-sectional manner. Additionally, the study characteristics of the only prospective study by Brandow et al., 2009 were fairly disparate from the present study (e.g., severe pain that required emergency room care, short time frame between completion of the PedsQL and follow-up measures). Using G*Power, we estimated that a sample size of 81 would be large enough to detect an effect of small to moderate size (f ² = .11) in the regressions (Faul, Erdfelder, Lang, & Buchner, 2007).

Results

Missing Data

Ten percent of caregivers who had HRQL ratings at Time 1 did not have ratings at Time 2. Simple t-tests and chi-square correlations were used to determine whether there were any statistically significant differences between caregivers and their children whose data were used in the study versus those with missing follow-up data. The following variables were examined: age, gender, history of neurobehavioral and medical concerns, insurance status, hematocrit, and HRQL ratings at Time 1. We examined sickle cell subtype descriptively. There were no statistically significant differences observed for any outcome. Children of caregivers with missing data had either the HbSS (n = 8) or HbSC (n = 1) sickle cell subtype.

Statistical Assumptions

All statistical assumptions for multiple hierarchical regression were met, with the exception of homoscedasticity for pain episode frequency and normality of residuals. Children’s HRQL scores were more variable at low levels of pain episodes and less variable as the number of pain episodes increased. This pattern may be the result of the distribution for pain episodes, which demonstrated a very large positive skew. A previous, large epidemiological study of pain in SCD demonstrated a similar distribution for pain episodes using a comparable measurement method, with most patients reporting few pain episodes requiring health care utilization (Platt et al., 1991). It is also possible that this pattern is specific to the present sample, who represent a diverse and relatively small group of families, or that the variability in time between ratings resulted in the nonnormal pain episode distribution. Conducting a square root transformation of the pain variable did not result in a substantial change to the distribution, so the original scaling was preserved.

Residual scores were substantially leptokurtic on all HRQL outcomes, which may indicate that additional variables not assessed in this study were influencing children’s HRQL scores over time. This study attempted to account for numerous variables that have been linked to HRQL outcomes in previous studies; however, few studies have examined which variables are linked to change in HRQL over time in children with SCD.

Casewise diagnostics were used to identify children whose scores significantly deviated from the sample and their effects on the models, which are referred to as influential cases. Children whose HRQL scores exceeded for the standardized DFBETAS values were evaluated further. These results are reported after the primary findings below under “Influential Cases.”

Multiple Hierarchical Regressions

Multiple hierarchical regressions were conducted to assess whether pain episode frequency between Time 1 and Time 2 was associated with changes in Physical, Psychosocial, and Total HRQL on the PedsQL. Table II provides average HRQL scores for caregiver reports at Time 1 and 2 for the sample. Bivariate correlations were used to identify potential confounding demographic and medical variables. Table III provides correlations for these variables, pain episode frequency, and changes in HRQL ratings (reported as standardized residuals). Covariates with correlations greater than or equal to .20 were retained in the regressions. The addition of interaction terms to the models did not account for a significant amount of additional variance and none of the individual β-coefficients for interactions were significant for any model. The interaction terms have been omitted from the final results. Hierarchical regression results containing each step of the model for HRQL can be found in Table IV to supplement the text below.

Table II.

Average Physical, Psychosocial, and Total Score HRQL Ratings at Time 1 and 2

HRQL outcome	Time 1 (M, SD)	Time 2 (M, SD)
Physical HRQL	65.28 (21.24)	62.25 (21.63)
Psychosocial HRQL	66.14 (17.33)	64.07 (15.25)
Total score HRQL	65.67 (16.72)	63.51 (15.69)

Note. HRQL are for 81 caregivers with ratings at both time points.

Table III.

Bivariate Correlations for Pain Episode Frequency, Covariates, and Change in HRQL

Variables	1	2	3	4	5	6	7	8	9	10
1. Physical HRQL	–	.48**	.68**	−.16	.05	.06	.17	−.20†	−.02	−.30**
2. Psychosocial HRQL		–	.96**	−.13	.11	.09	−.21†	−.12	−.03	−.30**
3. Total HRQL			–	−.20†	.07	.11	−.13	−.19	−.02	−.38**
4. Time interval				–	−.01	−.11	.19†	−.09	−.04	−.01
5. Gender					–	.03	.12	−.01	.08	.11
6. Age						–	−.16	.09	.19†	−.12
7. Insurance status							–	−.16	−.07	−.02
8. Neurobehavioral hx								–	−.06	.13
9. Medical hx									–	.16
10. Pain episodes (between T1 and T2)										–

Note. HRQL ratings are standardized residual scores derived from a bivariate regression that adjusted Time 2 (T2) ratings for Time 1 (T1). Correlations demonstrate the relationship between changes in HRQL from T1 to T2, pain, and demographic and medical covariates. *p < .05, **p < .01, ^†p < .10.

Table IV.

Multiple Hierarchical Regressions Predicting HRQL

Variable	β	ΔR2	Total R2
Physical HRQL
Step 1
Time 1 physical HRQL	.87	.76**	.76
Step 2
Time interval	−.09	.01	.77
Step 3
Neurobehavioral complications	−.10	.01	.78
Step 4
Pain episodes	−.15	.02*	.80
Psychosocial HRQL
Step 1
Time 1 psychosocial HRQL	.78	.60**	.60
Step 2
Time interval	−.08	.01	.61
Step 3
Insurance (code 1)	−.13	.02	.62
Insurance (code 2)	−.11
Step 4
Pain episodes	−.20	.04*	.66
Total HRQL
Step 1
Time 1 total HRQL	.90	.72**	.72
Step 2
Time interval	−.12	.01	.73
Step 3
Pain episodes	−.21	.04*	.78

Note. Table III provides abbreviated hierarchical regression results for HRQL outcomes to supplement those in the text. Insurance code 1 contrasted children with private insurance to those with Medicaid only and code 2 contrasted children with combined private insurance/Medicaid to those with Medicaid only. Results are provided for the variable that was added at each step of the regression. A bonferroni correction was used to adjust for multiple comparisons. *p < .016, **p < .001.

For Physical HRQL, the overall regression model was statistically significant, F(4,76) = 76.24, p < .001, accounting for 80% of the variance in Time 2 HRQL ratings. The majority (76%) of the variance was accounted for by Time 1 HRQL ratings, F(1,79) = 253.66, p < .001. As hypothesized, pain episodes accounted for a significant amount of additional variance, F(1,76) = 7.79, p = .007. Each additional pain episode resulted in a .94 decrease [SE = .34, 95% CI (−1.61 to −.27)] in Physical HRQL, as indicated by the unstandardized beta coefficient. Time interval, F(1,78) = 2.42, p = .124, and neurobehavioral complications, F(1,77) = 3.66, p = .059, did not contribute additional significant variance to the model.

For Psychosocial HRQL, the overall regression model was statistically significant, F(5,75) = 29.28, p < .001, accounting for 66% of the variance in Time 2 HRQL ratings. The majority (60%) of the variance was accounted for by Time 1 HRQL ratings, F(1,79) = 118.74, p < .001. As hypothesized, pain episodes accounted for a significant amount of additional variance, F(1,75) = 8.16, p = .006. Each additional pain episode resulted in a .88 decrease [SE = .31, 95% CI (−1.50 to −.27)] in Psychosocial HRQL. Time interval, F(1,78) = 1.35, p = .248, and insurance status, F(1,76) = 1.73, p = .185, did not contribute additional significant variance to the model.

Finally, for Total HRQL, the overall regression model was statistically significant, F(3,77) = 88.86, p < .001, accounting for 78% of the variance in Time 2 HRQL ratings. The majority (72%) of the variance was accounted for by Time 1 HRQL ratings, F(1,79) = 205.98, p < .001. As hypothesized, pain episodes accounted for a significant amount of additional variance, F(1,77) = 14.24, p = .001. Each additional pain episode resulted in a .96 decrease (SE = .25, 95% CI −1.46 to −.45) in Total HRQL. Time interval did not contribute additional significant variance to the model, F(1,78) = 3.43, p = .068. Figure 1 demonstrates the relationship between pain episode frequency and decreases in physical, psychosocial, and total HRQL over time.

Figure 1.

Pain episode frequency predicts decreases in caregiver-reported HRQL. All three graphs demonstrate the relationship between higher pain frequency and decreases in HRQL over time. Change in HRQL represents Time 2 HRQL ratings that have been adjusted for Time 1 ratings, represented in standardized residual scores (std. resid.).

Influential Cases

Casewise diagnostics revealed three to four influential cases per HRQL outcome. These cases were systematically deleted to determine their influence on the model. The deletion of these cases altered the statistical significance of neurobehavioral complications as a predictor of Physical HRQL and insurance status as a predictor of Psychosocial HRQL, with both variables moving towards statistically significance with the deletion of the outliers. Pain episodes continued to be a robust predictor of change in HRQL with these values deleted. Since pain was the primary predictor under consideration, the influential values were retained in the final models.

Discussion

This study used a prospective design to examine whether caregiver reports of the PedsQL were responsive to pain-related changes in HRQL in pediatric SCD. Pain episode frequency between Time 1 and Time 2 was found to be predictive of changes in physical, psychosocial, and overall HRQL, providing preliminary support for the responsiveness of the PedsQL to pain-related changes in HRQL over time in children with SCD.

The finding that caregiver proxy reports were responsive to changes in HRQL is encouraging. Caregivers may be asked to complete proxy reports of HRQL for several reasons, including when the child is too young or too ill to report on HRQL (Eiser & Morse, 2001). The finding that caregiver report was responsive to psychosocial HRQL is particularly notable given the inconsistencies found in previous studies with this domain in distinguishing between children with and without a history of pain and given concerns about the ability of caregivers to detect psychosocial functioning (Dampier et al., 2010; McClellan et al., 2008; Palermo et al., 2002; Panepinto et al., 2005, 2009). Although proxy reports cannot replace self-reports, caregiver perspectives are considered an integral part of HRQL assessment (Matza, Swensen, Flood, Secnik, & Leidy, 2004).

Change estimates in HRQL associated with each pain episode may be informative for those tracking changes in HRQL in children with SCD for clinical research or patient care. Caregiver-reported HRQL decreased by approximately one point (ranging from a .88 to .94 point decrease) per each additional pain episode experienced between time points. If replicated by additional studies, these change estimates may be useful to clinicians and interventionists in anticipating the amount of change in HRQL outcomes on the PedsQL that may occur in children whose pain episode frequency has changed over time. This information may be particularly helpful when combined with previously reported information on minimum clinically important differences (MCID) for the PedsQL, in order to determine when changes in HRQL warrant clinical attention or when changes signify an important change in response to treatment. For example, based on MCID estimates, approximately five episodes would be anticipated to result in a clinically meaningful drop in Total Score HRQL (Varni, Burwinkle, Seid, & Skarr, 2003). Future studies may also consider assessing subjective feelings about meaningful changes in the PedsQL that are specific to children with SCD.

The degree of change associated with pain episodes is much smaller than what was reported in Brandow et al. (2009)’s study of children admitted to the emergency room. These authors reported immediate, dramatic drops in both child- and caregiver-reported HRQL across all domains, with scores returning to those of SCD controls 1 week postdischarge. The discrepancies between this study and the present study are likely accounted for by differences in study methodology, including the population of children with SCD (i.e., children with severe pain requiring emergency room care versus children whose pain required any form of health care contact), the timing of the measurement (i.e., at the emergency room visit versus at routine health visits), and the examination of a single pain episode versus the cumulative effects of multiple episodes.

Covariates examined may warrant attention in future studies. In particular, history of neurobehavioral complications and insurance status emerged as possible predictors of change in HRQL. Despite the inclusion of these covariates, a high degree of variability in HRQL scores remained for children without any documented pain episodes, which may suggest that important variables were not included in the model. Insurance status, in particular, may have been problematic to use in the absence of additional measures of socioeconomic status, such as caregiver education and income, which have explained variability in HRQL in previous studies of children with SCD (Palermo et al., 2008; Panepinto et al., 2005, 2009). Additional covariates to consider in future studies include caregiver locus of control and neighborhood/community factors (Barakat et al., 2005; Palermo et al., 2008).

These findings should be interpreted within the context of limitations. First, pain episode frequency was measured using a health care utilization method, which despite its frequent use in research has certain limitations. Rather than being a pure measure of pain frequency, health care utilization measures can also capture aspects of the health care environment and experience (e.g., relationships with health care personnel) that may affect HRQL (Smith et al., 2005). Additionally, many pain episodes are cared for at home rather than via health care contact or visit and information from outside treating facilities may not be shared or documented in the medical chart (Shapiro et al., 1995); thus, the current pain measure likely underestimated the frequency of pain episodes experienced by some children in the study. Also, there was a great deal of variability in HRQL ratings for children with no documented pain episodes. It is possible that some of these children may have experienced pain at home that affected HRQL, but was not captured by our pain measure. The present study also did not capture features of pain aside from frequency, such as intensity, duration, and location of pain. Finally, this study included a wide age range and the timing between HRQL assessments was not standardized.

The generalizability of the results should also be considered. Compared to a previous, large study of HRQL in SCD with the PedsQL, average caregiver ratings in our sample were approximately five to nine points lower (Dampier et al., 2010). The children within the catchment area of our hospital-based clinic likely had higher clinical morbidity than is typical of the overall population of children with SCD, resulting in lower HRQL scores. Additionally, families in our study are likely to have higher incomes than would be found in families who receive services in other settings, such as local health departments. These considerations are particularly important when evaluating the magnitude of changes in HRQL observed because these estimates may be specific to our sample (Terwee et al., 2003). The findings may also not generalize to children whose pain is predominantly cared for through home management rather than health care contact. Finally, we had approximately 10% missing follow-up data for caregivers. Although no statistically significant differences were found among caregivers with and without follow-up data, we may not have had statistical power to detect these differences.

In conclusion, this study provides preliminary support for the use of caregiver report of the PedsQL to assess changes in HRQL due to pain episode frequency in pediatric SCD. As noted by Terwee et al. (2003), the responsiveness to changes in specific aspects of health is a critical component of a measure’s performance that is frequently overlooked in psychometric studies. This information has the potential to guide the use of these measures to monitor patient progress and response to treatment for pain from the child and caregiver’s perspective. Establishing the measurement properties of the PedsQL in SCD is particularly crucial, given the absence of disease-specific measures of HRQL in SCD and the need for efficient measurement tools that can be used in medical settings (McClellan et al., 2008). Future studies should consider replicating the effect sizes observed for change in HRQL, assessing the responsiveness of the PedsQL child report to pain, and determining the responsiveness of the PedsQL to multiple pain features.

Funding

This work was supported by the National Institutes of Health, National Institute of General Medical Sciences (T32 GM081740 to A.M.S.); an intramural award from the Vice President for Research at the University of South Carolina (to J.S.); and the Pfizer Fellowship in Health Disparities (to C.B.M.).

Conflicts of interest: None declared.